Almost every manufacturer that sells through distribution measures its channel the same way. Once a year, a survey goes out to the distributors that carry the line. It asks how satisfied they are, how likely they would be to recommend, how the relationship is going. The results come back, they are averaged, they go into a slide, and the slide says the number went up or down by some amount that nobody can act on.
The problem is not that the survey is badly written. The problem is that it measures the wrong object.
Sentiment and behavior are different things, and only one of them pays
A distributor can rate a manufacturer a promoter and still leave that brand’s line on the shelf. There is nothing contradictory about it. The relationship genuinely is fine. The rep is pleasant, the rebate cleared, nobody has had a fight. And when a customer calls in and asks for something that three manufacturers make, inside sales quotes a different box, because that box is on the shelf, or the branch manager came up on it, or the last warranty claim on yours turned into an argument that the person quoting it personally had to sit through.
Recommendation is a stated intention about a hypothetical. Reaching for the box is a behavior under time pressure with a customer standing there. The gap between those two is where a line is won or lost, and a question about willingness to recommend cannot see into it, because the person answering is not lying and does not know they are giving you a useless answer.
This is also why channel scores drift upward over time and then a line dies anyway. Satisfaction is easy to move. You can move it with a golf outing. Behavior on the quote moves only when one of the underlying conditions changes, and the conditions are specific: how hard you are to order from, whether the margin survives a quoted job, who answers when an application question lands on a Friday afternoon, whether your product arrives as specified, and whether their own people lead with you when nobody has asked for you by name.
That last one is the whole game, and it is the one a generic instrument never reaches.
Five conditions, not one number
An instrument that is going to be useful has to score the conditions separately, because they have different owners and different fixes.
Ease of doing business is an operations problem. Pricing is a commercial policy problem. Support is a staffing and capability problem. Quality is a manufacturing problem. Internal promotion is the consequence of the other four plus your positioning, and it is the only one that predicts which brand gets quoted.
Collapsing those into a single index number is what makes channel research feel useless. You are handed a 7.2 and no idea which lever moves it, so the meeting turns into a debate about whether 7.2 is good, which is a debate nobody can win or lose. Five scores, each in the language distributors actually use, give you an argument about a specific thing that a specific person can change by a specific date.
Blending voices destroys the finding
Most channel surveys go to one contact per distributor, usually the principal or the person who signs the agreement. That is the wrong sample, and it is wrong in a predictable direction.
The owner, the outside salesperson, and the inside sales team do not experience your line the same way and do not disagree at random. The owner sees the annual number, the rebate, and the relationship with your regional manager. The outside salesperson sees whether you make them look competent in front of a customer. Inside sales sees what is on the shelf and what the customer asks for. A manufacturer can be genuinely good to the owner and invisible to inside sales, and if you only asked the owner you will never learn that.
Averaging those three answers is worse than asking only one of them, because the average produces a figure that is true of nobody and looks like a measurement. A 6.8 that is really a 9 from the owner and a 4 from inside sales is not a mediocre relationship. It is a specific, fixable, expensive problem wearing the costume of a mediocre relationship.
So the roles get scored separately and reported separately, and where they disagree the disagreement is reported as the finding rather than resolved into a single figure.
The anonymity problem, which is the reason most of this is never attempted
Here is the constraint that makes channel research genuinely hard, and the reason most manufacturers settle for the satisfaction survey they know is weak.
A useful report has to name the distributor. “Internal promotion is weak in your network” is not actionable. “Internal promotion is weak at these four distributors and strong at these three” is a decision about where market development money goes next year. But the moment the report names the distributor, the person answering inside that distributor is exposed, and they know it, and they will tell you support is fine.
The instrument therefore has to be designed around the protection, not have the protection bolted on afterward. What that requires in practice:
Responses are anonymized by role within the distributor, so the card says what inside sales said without saying which person said it. No card is issued at all for a distributor with fewer than three respondents, because at two you can identify people by elimination. When a distributor comes in thin, the shortfall is reported as a finding rather than papered over, since a distributor that will not put three people on the phone has told you something. Verbatim quotes are cleared individually before they are attributed to anything. And every one of those rules is stated to the respondent at recruitment, because a promise made after the fact is not a promise.
Skip this and you have paid a considerable amount of money for a mirror.
Benchmarks, and the honest version of a small pool
A manufacturer’s first question about any scored instrument is what a good score looks like. The honest answer, at the start, is that the useful comparison is not against an industry figure at all.
Your first study scores your distributors against each other. That comparison is immediately actionable, because the same line, the same pricing policy, and the same regional manager are constant across the network, so a distributor that scores low on internal promotion while its neighbors score high is a fact about that distributor rather than about your category. Cross-industry benchmarks are the weakest kind of comparison anyway, because a distributor’s willingness to lead with a line depends on the category, the local competitive set, and how many lines the distributor already carries.
Comparison against an outside pool becomes worth having in year two, and it becomes worth more each year, which is why any instrument like this has to keep the right to publish de-identified aggregate findings. It is also why any report quoting a benchmark should state the number of studies behind it. A benchmark offered without its n is the same trick as the blended average: a figure that sounds like a measurement and is not one.
What it has to survive
The test of a channel instrument is not whether it is methodologically defensible in the abstract. It is whether it survives a line review.
That means it has to arrive before the decision it is supposed to inform, not a quarter after. It has to name distributors, because the decision is about distributors. It has to sort them into things you can actually do, which in practice is three lists: protect, fix, and replace or constrain. And it has to separate a company problem from a distributor problem, because internal promotion weak everywhere is your positioning and internal promotion weak in three branches out of twenty is one local manager, and confusing those two is how a training budget gets spent on the wrong continent.
None of that is exotic. It is what you would build if you started from the question a vice president of sales actually has, which is not how do they feel about us.
It is whether their people will pull our line tomorrow morning without being asked, and if not, which reason is in the way.
The instrument this argument produced. Channel Health Index